Options
Acoustic correlates of accentedness and comprehensibility in checked syllables of L2 Cantonese by Mandarin learners
Author(s)
Date Issued
2024
Citation
Yang, Y., Zou, Y., & Han, D. (8 December 2024). Acoustic correlates of accentedness and comprehensibility in checked syllables of L2 Cantonese by Mandarin learners. 2024 Annual Research Forum of the Linguistic Society of Hong Kong, Hong Kong Shue Yan University.
Type
Conference Paper
Abstract
Purpose: One distinctive feature of Cantonese is its checked syllables, which are closed
syllables ending with voiceless plosives /-p/, /-t/ or /-k/ (Wong & Chan, 2018). Among the
literature on second language (L2) Cantonese speech learning, little attention has focused on
these checked syllables. This study aimed to address this gap by investigating the acoustic
correlates of accentedness and comprehensibility judgements in L2 Cantonese checked
syllables produced by Mandarin-speaking learners. Lacking checked syllables, Mandarin
presents a unique linguistic contrast, making this an ideal case to examine the acquisition of
Cantonese checked syllables.
Methods: Forty Mandarin-speaking learners of Cantonese (17 males and 23 females; aged:
29.81 ± 5.77) were invited to record the numbers from zero to ten in a speech production
experiment. One clearly pronounced token for each of the four checked syllables, ‘six luk6’,
‘seven cat1’, ‘eight baat3’, ‘ten sap6’ was manually extracted from the recording of each
speaker. These 160 tokens (4 syllables x 40 speakers) were normalised for intensity and
included as the stimuli in this study. Thirty-two native Cantonese speakers (9 males and 23
females; aged: 20.27 ± 2.29), all novice listeners (Suzuki & Kormos, 2020), were recruited to
rate the accentedness and comprehensibility of each trial on 9-point Likert scales. Following
previous research (Chai & Ye, 2022; Oakden, 2017; Yu et al., 2022), this study measured vowel
quality (frequencies of the first and second formants, F1 and F2), pitch contours (discrete
cosine transform values, DCT1, DCT2 and DCT3), duration and voice quality (jitter).
Generalised additive models (GAMs) were constructed to examine how the acoustic cues
explained the perceived accentedness and comprehensibility by native listeners.
Summary of findings: For each rater, the correlation coefficients between accentedness and
comprehensibility ratings ranged from 0.445 to 0.953 (ps < 0.001). The GAM models showed
that all the seven acoustic cues had significant effects for perceived accentedness and
comprehensibility for the syllables ‘six luk6’ and ‘ten sap6’, suggesting that all these cues
might have contributed to the judgement for Tone 6, the low-level tone. For the syllable ‘seven
cat1’, all seven cues were effective for accentedness ratings, and six out of seven cues (all but
duration) were effective for comprehensibility ratings. For the syllable ‘eight baat3’, only five
cues had significant effects for accentedness and four had significant effects for
comprehensibility, revealing a more complex picture for this syllable, which might have been
caused by its vowel length as Mandarin does not have long vowels. Overall, F2, DCT2 and
jitter seemed to be reliable cues for predicting the ratings as they were effective in all the eight
models. The results reveal acoustic correlates influencing native speaker judgments, offering
valuable insights for developing training paradigms to enhance L2 Cantonese speech.
Availability at HKSYU Library
